The Roles of Modified Input and Output in The Incidental Acquisition of Word Meanings

SSLA, 21, 285301. Printed in the United States of America.
THE ROLES OF MODIFIED INPUT

AND OUTPUT IN THE
INCIDENTAL ACQUISITION
OF WORD MEANINGS
Rod Ellis
University of Auckland
Xien He
Temple University
This article reports an experimental study of the differential effects of

premodified input, interactionally modified input, and modified output
on the comprehension of directions in a listen-and-do task and the
acquisition of new words embedded in the directions. The modified
output group achieved higher comprehension and vocabulary acqui-
sition scores than either of the input groups. There was no difference
between the premodified and interactionally modified input groups.
The advantage of the modified output group is explained in terms
of the qualitatively superior dialogic interaction that occurred in this
condition rather than in terms of actual language production.
According to the Input Hypothesis (Krashen, 1985), learners acquire an L2

when they are exposed to comprehensible input that contains linguistic forms
slightly in advance of their current interlanguage system (i + 1). Krashen has
consistently refused to countenance any role for learner output in acquisition,
other than that of providing auto-input. For example, Krashen (1994) claims
that only comprehensible input is consistently effective in increasing profi-
ciency and that more skill-building, more correction, and more output do
Address correspondence to Rod Ellis, Director, Institute of Language Teaching and Learning, Faculty
of Arts, The University of Auckland, Private Bag 92019, Auckland, New Zealand; e-mail: r.ellis@
auckland.ac.nz.
1999 Cambridge University Press 0272-2631/99 $9.50 285
Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 17 Jul 2017 at 13:56:21, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://www.cambridge.org/core/product/3B026C16F7E59E3458121D43E055741D
286 Rod Ellis and Xien He
not consistently result in more proficiency (p. 48). For Krashen, the ability to
produce the L2 is the result, not the cause, of acquisition.
Like the Input Hypothesis, the Interaction Hypothesis (Long, 1981) initially
viewed input rather than output as the source of acquisition. However, unlike
the Input Hypothesis, it did allow a constitutive role for learner output. Long
argued that, when learners negotiate meaning by means of requests for clarifi-
cation or confirmation checks, they can obtain interactionally modified input
that both helps them to comprehend the input and focuses their attention on
new or partially learned linguistic forms, thus enabling their acquisition. Pro-
tocol (1) below illustrates how meaning negotiation can create the conditions
for a learner to acquire a new word, lampshade.
(1) T: Please put the lampshade on the desk.

S: What is a lampshade?
T: A lampshade is placed round or over a lamp.
According to the early version of the Interaction Hypothesis, then, learner out-
put facilitates acquisition when it elicits modified input.
In later versions of the Interaction Hypothesis, however, Long has recog-
nized that meaning negotiation can induce learners to modify their own out-
put and that this, too, may promote acquisition. Pica (1994) and Long (1996)
argue that, when learners receive implicit negative feedback on their attempts
to communicate, they may attempt to reformulate their initial utterances,
thereby promoting acquisition. Below is an example of output modification:
(2) S1: Please put the comforter on the bed.

S2: Comfortable?
S1: No, comforter. Comforter is like a blanket. Do you know meaning?
S2: Yes. Put comforter on the bed?
It is clear that such exchanges afford more than comprehensible input; they
also allow learners to rehearse new items (such as comforter) in production.
Swain (1985) argues that when learners are required to produce pushed out-
put (i.e., output that is precise, coherent, and appropriate) they may be forced
to move from semantic processing to syntactic processing. She notes that
learners can comprehend an utterance by utilizing top-down strategies involv-
ing existing knowledge and context but that to produce pushed output they
need to pay attention to the means of expression. Swain (1995) identifies a
number of different roles for output: It enables learners to improve their flu-
ency, it may help learners to recognize a gap between what they want to say
and what they can say, it serves as a means by which learners can test
hypotheses about comprehensibility or linguistic correctness, and it can help
learners to develop metalinguistic knowledge of how the L2 works. Swains po-
sition is that both comprehensible input and comprehensible output are im-
portant for L2 acquisition.
The Roles of Modified Input and Output 287
Most of the research to date has focused on the effects of modified input
on comprehension. As Pica (1994) pointed out:
researchers have given less attention to identifying a direct impact for

learners negotiation on restructuring of their interlanguage grammar than
to documenting the contributions of negotiation in bringing about condi-
tions claimed to be helpful for SLA, namely learners comprehension of L2
input, their production of modified output, and their attention to L2 form.
(pp. 499500)
Studies by Pica, Young, and Doughty (1987), Loschky (1994), and Ellis, Ta-
naka, and Yamazaki (1994) have compared how (a) baseline input (i.e., the
kind of input native speakers hear when listening to other native speakers),
(b) premodified input (i.e., input that has been simplified by making it more
redundant and less grammatically complex), and (c) interactionally modified
input (i.e., input that has been modified as a result of meaning negotiation)
affect adult learners comprehension. These studies all found that interaction-
ally modified input worked best. However, this kind of input seems less effec-
tive where young children are concerned. Ellis and Heimbach (1997) found
that 56-year-old children were reluctant to negotiate directions with their
teacher and, as a result, achieved low levels of comprehension.
A few studies have investigated the relative effects of interactionally modi-
fied and premodified input on L2 acquisition, with mixed results. Loschky
(1994) found no effect; baseline, premodified, and interactionally modified
groups performed at the same level on a sentence verification test and a vo-
cabulary test. In contrast, Ellis et al. (1994) found a very clear effect for type
of input. The interactionally modified group acquired more new words than
the premodified group, which in turn acquired more than the baseline group
in a posttest administered immediately after the treatment.1 A possible expla-
nation for the different results obtained by these studies is that, whereas
Loschky presented all groups with a written list of the words to be used in his
tasks, Ellis et al. did not.
The clearest evidence that interaction assists acquisition comes from
Mackey (1995). Mackey compared the effects of interactionally modified input
and premodified input on L2 learners acquisition of English question pat-
terns, basing her research on the developmental stages that have been ob-
served in question formation. She found that groups of learners who
participated in interaction showed greater developmental gains in producing
questions than did both a group that merely observed other learners interact-
ing and a group that just heard scripted premodified input.2 That is, they ad-
vanced through more developmental stages and also produced significantly
more higher-level question structures. However, the increase in the produc-
tion of higher-level structures only emerged in a posttest administered one
week later, indicating that interactionally modified input may have a delayed
rather than an immediate effect on acquisition.
A problem arises in interpreting the results of these studies both with re-
gard to comprehension and acquisition. The tasks that supplied interaction-

ally modified input took longer than those based on premodified input. We
cannot tell, therefore, whether interactionally modified input works best be-
cause it enables learners to sort out misunderstandings and construct a
shared mental model of the task at hand, as suggested by Brown (1995), or
because learners have more time to process input.
Ellis (1995) analyzed the modified input from one of the studies reported
by Ellis et al. (1994). He was able to show that, although interactionally modi-
fied input led to higher levels of vocabulary acquisition than premodified in-
put, overall it was less efficient (i.e., premodified input worked much better in
terms of words acquired per minute of input). Ellis also investigated what it
was in the input that enabled the learners to learn some of the new words
more easily than others. In the case of premodified input, he found that learn-
ers remembered best those words that occurred in many different directions
and in longer directions. In the case of interactionally modified input, range
also proved important but the analysis showed that learners were less likely
to remember words when the negotiation led to very long definitions.3
There has also been very little research that has investigated the effects of
learner output on acquisition. Pica et al. (1989) showed that learners were
more likely to modify their output by making it more grammatical in response
to requests for clarification. However, because of the greater number of con-
firmation checks, these accounted for the greatest number of output modifica-
tions overall. In a small-scale study, Nobuyoshi and Ellis (1993) found that
some learners who were pushed to reformulate by means of requests for clari-
fications in a storytelling task responded by correcting their past-tense errors
and subsequently used this feature more accurately when they repeated the
task one week later.
To sum up, SLA theories differ in which learning conditions they claim are
optimal for L2 acquisition. Different theoretical arguments have been ad-
vanced on behalf of premodified input, interactionally modified input, and
modified output. As yet, however, few studies have tested the claims of these
theories and the results to date have been mixed.
RESEARCH QUESTIONS
The study described below was designed to investigate the effects of various
conditions of exposure and use of new L2 words on L2 learners comprehen-
sion of input containing the words and their acquisition of these words. The
research questions were:
1. What are the relative effects of premodified input, interactionally modified input,
and modified output on L2 learners comprehension of directions containing new
L2 words?
and modified output on L2 learners ability to subsequently recognize new L2
words?
and modified output on L2 learners ability to subsequently produce new L2
words?
METHOD
Subjects
The subjects of this study were 50 students from six intermediate-level classes
of the Intensive English Language Program (IELP) at Temple University in Phil-
adelphia. Twenty-seven of the subjects were male and 23 were female. A ma-
jority were Korean (26), 15 came from other parts of Asia (Japan, Taiwan,
Hong Kong, Malaysia, and Thailand), and 12 from other parts of the world.
Their length of stay in the United States prior to the study varied from 10 days
to 2 years. Almost all of the subjects had studied English in their home coun-
tries for a period of at least 5 years. Their ages ranged from 18 to 44 years.
The teacher who taught all three lessons that composed the treatments in
this study was the assistant director of the IELP program. She was a very ex-
perienced teacher and familiar with the kinds of students in the classes. How-
ever, she had not taught any of the classes previously.
Design
A multifactorial design with intact groups was used in this study. Existing
classes in the IELP were designated as the Premodified Group (n = 18), the In-
teractionally Modified Group (n = 16), and the Output Group (n = 16). There
was no control group.4 To establish whether there were any differences in gen-
eral language proficiency in the subjects in the three groups, the Secondary
Level English Proficiency Test (SLEP) was administered a few weeks before
the commencement of the study. A one-way analysis of variance (ANOVA)
showed that the difference in the groups scores was not statistically signifi-
cant, F = 2.26, p > .05.
The three groups each completed the following: (a) a pretest, administered
one week before the treatment, (b) the treatment, which varied for the three
groups, (c) Posttest 1, administered one week after the treatment, (d) Posttest
2, administered 2 weeks after the treatment, (e) Posttest 3, administered 3
weeks after the treatment, (f) Posttest 4, administered 4 weeks after the treat-
ment, and (g) Posttest 5, administered at the same time as Posttest 4.
Instruments
The Pretest. The subjects were asked to read a list of 50 English words and
check each word they recognized. In addition to items labeling furniture, the
list included distracting items (e.g., pizza and lemon). The 10 items least
known by the subjects were selected for the study. The overall level of non-
recognition of these items was 88%. The target items all referred to furniture.
They were: recliner, lampshade, comforter, stool, chest of drawers, china cabinet,
wardrobe, rocker, dresser, and cushion.
Posttest 1. A picture-matching test was designed to measure the subjects
ability to recognize the meanings of the target words. The subjects were given
individual pictures of pieces of furniture and a list of the 10 target words.
They were asked to match the pictures and the words. They were awarded
one mark for each picture they labeled correctly (total = 10).
Posttest 2. This test was designed to measure the ability of the subjects to
produce the target words orally. The subjects were put in pairs. Each subject
was given a matrix picture of an apartment and 10 pictures showing individual
pieces of furniture. The researcher quickly read out the words labeling the
pictures. Subjects were required to make up a direction instructing where
they wanted each piece of furniture to be placed in the matrix picture of the
apartment. They then gave each direction orally to their partners, who tried
to carry it out by writing in the number of the picture in the correct position
in the matrix picture. Subjects were allowed to negotiate meaning if they did
not understand their partners directions. After completing this task, the stu-
dents were given pictures of the individual pieces of furniture and asked to
label them from memory. They were awarded one mark for each picture they
labeled correctly (total = 10). In effect, this test replicated the treatment for
the Output Group (see below). It was administered to 10 subjects in the Pre-
modified Group, 8 in the Interactionally Modified Group, and 9 in the Output
Group. In this way, it was still possible to measure the effects of the original
treatment on scores obtained in Posttests 3 and 4 by examining only those
subjects who did not complete Posttest 2.
Posttest 3. This test was the same as Posttest 1. The orders of the pictures
and the words were changed, however.
Posttest 4. This test required subjects to label pictures of the individual
pieces of furniture, as in Posttest 2.
Posttest 5. This was the same as Posttest 1 (i.e., subjects completed the
picture-matching test for the third time).
Treatment
The treatments provided to all three groups involved the use of the same task
performed under different conditions. The subjects were given a matrix pic-
ture of an apartment and a series of small pictures showing individual pieces
of furniture. They were asked to listen to or to produce directions that gave
instructions about where to place the pieces of furniture in the matrix picture
of the apartment. The task contained an inbuilt measure of the subjects com-
prehension of the directions (i.e., whether they had written the numbers of
the correct pictures in the correct position in the matrix picture). Therefore,
no separate measure of comprehension was required.
The directions contained words that the pretest had shown were new to
the majority of students. The task required the learners to attend primarily
to meaning but was designed in such a way that learners attention was drawn
to specific linguistic items that constituted the learning targets. However,
such a task does not require learners to learn the items intentionally; the
learning of the items is a possible natural outcome of performing the task. In
this respect, the task catered to incidental rather than intentional language
learning.5
The conditions under which this task was performed in the three experi-
mental groups are described below.
The Premodified Input Treatment. Initially, baseline directions were ob-
tained by asking two native speakers to make up directions about where to
place each of the pieces of furniture in the matrix picture of the apartment.
These were recorded and transcribed. Here is an example of the kind of direc-
tion they produced:
(3) Find the sofa. Id like you to put the cushion on the sofa.
Next, students drawn from the same population as the subjects in the
study listened to the baseline directions and negotiated meaning when they
did not understand. The interactions were audio recorded and used to pre-
pare the premodified directions. An example of a premodified direction is:
(4) Find the cushion. And Id like you to put the cushion on the sofa. A cushion is like a
pad and you put it on the sofa to support your back.
Ten such directions were prepared, one for each target item.
The teacher gave out the matrix picture of the apartment, on which the
subjects wrote their names, and the small pictures of the different pieces of
furniture, which she very quickly named. She then read out each direction at a
slow rate (approximately 90 words per minute). The students listened to each
direction, chose from the pictures of the pieces of furniture, and wrote the
number of the picture they had chosen in the matrix picture of the apartment.
The students were not allowed to interact with the teacher. To ensure that
this treatment took 45 minutes, the teacher went through the directions a sec-
ond time. At the end of the lesson, the teacher collected each subjects com-
pleted matrix picture.
The Interactionally Modified Input Treatment. Baseline directions were
used in this treatment. The teacher wrote a number of formulae for requesting
clarification on the board (e.g., What is a ? and Could you say it again?).
She then read out each baseline direction at a normal speed (about 180 words
per minute). The students were encouraged to interact with the teacher using
the requesting formulae and she attempted to help them understand by add-
ing definitional information, as in example (5).
(5) T: Here is a cushion. Would you please put the cushion on the sofa?
S: What is a cushion?
T: A cushion is a small bag filled with soft stuff.
The teacher gave out the matrix picture of the apartment and the small
pictures depicting pieces of furniture. After the students had negotiated each
baseline direction, they chose one of the small pictures and wrote its number
in the matrix picture. The teacher went through the baseline directions a sec-
ond time to ensure that the lesson lasted 45 minutes. The completed matrix
pictures were collected at the end of the treatment.
The Negotiated Output Treatment. The teacher gave out the matrix picture
of the apartment, the small pictures of pieces of furniture, and a blank piece
of paper. She then read out the words labeling the small pictures and the sub-
jects wrote down each word next to the corresponding picture. The teacher
then gave an example of how to write a direction. Each student then wrote 10
directions, one for each small picture. When they were finished, the teacher
put the students into pairs and asked them to exchange their directions
orally. They then performed the task in the same way as the other groups (i.e.,
they wrote the numbers of the small pictures in the matrix picture in accor-
dance with the directions given), negotiating meaning if they did not under-
stand a direction. This afforded opportunities for the students to modify their
own output. Each student, in turn, gave 10 directions. At the end the teacher
collected the matrix pictures. This treatment took 45 minutes.
RESULTS
In this study the independent variable was the kind of treatment that each
group received (premodified input, interactionally modified input, and modi-
fied output). There were three dependent variables:
1. Comprehension: This was determined by inspecting the subjects completed ma-

trix pictures to see how many of the directions they had understood. The maxi-
mum score was 10.
2. Vocabulary acquisition (recognition): Vocabulary recognition scores were ob-
tained on three separate occasions from the picture-matching test (i.e., Posttest 1,
Posttest 3, and Posttest 5). The maximum score was 10.
3. Vocabulary acquisition (production): Vocabulary production scores were obtained
on two separate occasions from the picture-labeling test (i.e., Posttest 2 and Post-
test 4). The maximum score was 10.
To determine whether there were any overall differences among the treat-
ment groups, a multiple analysis of variance (MANOVA) and univariate F tests
Table 1. Descriptive statistics for the

comprehension scores
Group
Premodified Interactionally Modified

Measure input modified output
n 18 16 16
M 6.78 7.13 8.12
SD 1.52 0.81 0.89
Minimum 3 6 6
Maximum 9 8 9
Table 2. Scheffe test of differences in comprehension scores

among the three groups
Group

Group input modified output
Premodified input 0.35 1.35**

Interactionally modified 1.00*
Note. *p < .05, **p < .01.
were performed using the comprehension scores and the scores for Posttests
1, 3, 4, and 5. The scores for Posttest 2 were not included in these analyses as
this test was administered to only half the sample. The MANOVA realized an
overall statistically significant difference, Wilkss Lambda F = 2.46, p > .02. The
univariate F tests showed significant group differences for all three dependent
variables. Descriptive statistics and the results of separate analyses of vari-
ance (ANOVAs) are reported below for comprehension, vocabulary acquisi-
tion (recognition), and vocabulary acquisition (production).
Comprehension
The premodified input group scored a mean of 6.67 out of 10, the interaction-
ally modified group 7.13, and the modified output group 8.13 (see Table 1).
There was an overall statistically significant difference among the groups, F =
6.31, p < .01. A post hoc Scheffe test showed that the modified output group
scored significantly higher than both the other groups (see Table 2). However,
the difference between the premodified and interactionally modified groups
was not statistically significant.
Vocabulary Acquisition (Recognition)

In Posttest 1, the premodified group scored a mean of 6.17, the interactionally
modified group 7.0, and the modified output group 8.19 (see Table 3). An
Table 3. Descriptive statistics for the three vocabulary

recognition posttests
Group
Premodified input Interactionally Modified output

(n = 18) modified (n = 16) (n = 16)
Measure PT1 PT3 PT5 PT1 PT3 PT5 PT1 PT3 PT5
M 6.17 6.33 6.72 7.00 7.56 7.31 8.19 9.00 8.63
SD 1.98 2.11 1.84 1.32 1.71 1.35 0.98 1.26 1.26
Minimum 2 3 2 5 4 6 6 7 6
Maximum 8 9 9 9 10 10 10 10 10
Note. PT1 = Posttest 1, PT3 = Posttest 3, PT5 = Posttest 5.
Table 4. Scheffe test of differences in vocabulary recognition

scores (Posttest 1) among the three groups
Group


Note. *p < .05, **p < .01.

Group

Premodified input 1.23* 2.67**

Note. *p < .05, **p < .01.
ANOVA showed these differences to be statistically significant, F = 7.63, p <

.05. A post hoc Scheffe test indicated that the modified output group signifi-
cantly outscored both the premodified input and the interactionally modified
groups (see Table 4). Again, the difference between the two input groups was
not statistically significant.
The results for Posttest 3 differed slightly (see Table 3). Again, there was a
significant overall difference in the groups, as revealed by the ANOVA, F =
9.84, p < .05. Also, the post hoc Scheffe test again showed that the modified
output group significantly outscored the other two groups (see Table 5). How-
ever, the interactionally modified group also scored significantly higher than

Group


Note. *p < .05, **p < .01.
Table 7. Descriptive statistics for the two vocabulary

production posttests
Group

input modified output
Measure PT2 PT4 PT2 PT4 PT2 PT4
n 10 18 8 16 9 16
M 5.60 6.22 5.75 6.44 7.55 8.19
SD 1.08 2.34 1.39 1.93 1.01 1.11
Minimum 4 1 3 2 6 6
Maximum 7 9 7 9 9 9
Note. PT2 = Posttest 2, PT4 = Posttest 4.
the premodified group. In other words, both the modified output group and
the interactionally modified input group did better than the premodified in-
put group.
In the final vocabulary recognition test (Posttest 5), administered 4 weeks
after the initial treatment, the results resembled those for Posttest 1. That is,
the ANOVA showed a significant overall difference among the groups, F = 6.84,
p < .05, but the post hoc Scheffe test indicated that, whereas the modified out-
put group again outperformed the other two groups, the difference in scores
obtained by the two input groups was not statistically different (see Table 6).
Vocabulary Acquisition (Production)

The ability of the subjects to produce the target lexical items was measured
by Posttest 2 and Posttest 4. Only a subset of the total sample completed Post-
test 2. The descriptive statistics are shown in Table 7. An ANOVA demon-
strated that there were significant differences among the three groups, F =
8.03, p < .01. A post hoc Scheffe test indicated that the modified output group
was able to produce significantly more of the target items than either the pre-

Group


Note. *p < .05, **p < .01.

Group


Note. *p < .05, **p < .01.
modified or the interactionally modified input groups (see Table 8). However,
the difference in scores for the two input groups was not statistically signifi-
cant.
In the case of Posttest 4, results for the total sample are reported. The de-
scriptive statistics can be seen in Table 7. Again, the ANOVA revealed signifi-
cant differences among the three groups, F = 6.45, p < .01, and again a post hoc
Scheffe test showed that the modified output group outperformed both the
input groups, but that the input groups produced the target words at a similar
level (see Table 9).
Summary
Overall these results are very consistent. The group that had had the opportu-
nity to produce and modify their own directions outperformed the two groups
that had experienced only modified input in their ability to comprehend the
directions, to recognize the new lexical items in picture-matching tests, and to
produce the items in picture-labeling tests. This pattern of results was ob-
served immediately after the treatment and at various times up to 4 weeks
later.6 In contrast, although the group that had received interactionally modi-
fied input consistently outscored the group that had heard only premodified
input in comprehension and on the vocabulary tests, the differences were
small and reached significance only in the case of one of the posttests (Post-
test 3: vocabulary recognition).
DISCUSSION
The first research question asked about the relative effects of premodified in-
put, interactionally modified input, and modified output on L2 learners com-
prehension. The results of this study indicate that reasonable levels of
comprehension can be achieved in all three conditions. Learners in the pre-
modified input group had a mean comprehension score of 68%, those in the
interactionally modified input group 71%, and those in the modified output
group 81%. These results suggest that the kind of listen-and-do tasks used in
this study can lead to high levels of comprehension even when they involve
unfamiliar words. The results also show that the modified output condition
worked better than the input conditions. However, in contrast to previous
studies, this study failed to demonstrate any advantage for interactionally
modified input over premodified input, possibly because this study controlled
for the time taken to complete each task. Thus, the results suggest that the
activity resulting from the input tasks was very similar, whereas that derived
from the modified output task was notably different.
The second research question addressed the effects of the various task
conditions on learners ability to subsequently recognize the target items.
Here again the results are encouraging; the learners in all three conditions
manifested high levels of acquisition (between 62% and 90%) and, further-
more, there was clear evidence that they were able to remember the words
over time. The modified output condition proved superior to either of the in-
put conditions in promoting retention. There is also some evidence that inter-
actionally modified input was more beneficial than premodified input.
However, the difference in scores between the two groups was slight and only
achieved significance in one of the posttests. This result again contradicts the
findings of the Ellis et al. (1994) study, which demonstrated a clear advantage
for interactionally modified input in a recognition posttest. Again, however,
this can be explained by the fact that in this study both input groups received
the same amount of time to complete their tasks and that the tasks involved
similar activities.
The third research question addressed the effects of the various task con-
ditions on learners ability to subsequently produce the target items. As was
to be expected, the scores on the production tests were lower than those on
the recognition tests but were still very respectable (between 52% and 82%).
The comparative effects of the different task conditions were very similar to
those for comprehension and vocabulary recognition; the modified output
group outperformed the other two groups, whereas the difference in scores
between the premodified input and interactionally modified input groups was
small and did not reach statistical significance.
Why did the modified output group consistently outperform the two input
groups in comprehension and vocabulary acquisition? The obvious explana-
tion is that producing new words helps learners to process them more deeply,
whether for comprehension or for acquisition, than simply hearing them.
However, we feel that there is more to it than this. We believe that the modi-
fied output condition afforded the learners a qualitatively different discourse
experience. To illustrate this, let us consider the three protocols below.
(6) T: Here is a rocker. Please put the rocker next to the sofa. A rocker is a chair which
can be rocked back and forth.
(7) T: Here is a rocker. Please put the rocker next to the sofa in the living room.
S: What is the rocker?
T: A rocker is a chair that can be rocked back and forth.
S: One more time.
T: A rocker is a chair which can be rocked back and forth.
S: Please repeat.
T: A rocker is a chair that can be rocked back and forth.
(8) S1: Please put the rocker on the living room.
S2: What is rocker?
S1: Rocker is like chair. You can sit and move. Look at the picture. You know now?
S2: Yes. Put rocker where?
S1: In the living room. There are three rooms in your big picture. Put rocker in the
room in the middle of the picture.
S2: OK.
The three groups comprehension scores for this item were 77%, 75%, and
93%. Similar differences were evident in the various vocabulary tests. It is not
difficult to see why the modified output group outscored the other two
groups. Protocol (8) is qualitatively different from (6) and (7) in a number of
respects. First, the definitional information provided by S1 is couched in vo-
cabulary (chair, sit, and move) that is likely to be familiar to S2. In contrast,
the definition supplied by the teacher in (6) and (7) uses a low-frequency item
(rock) that the students may not have known. Second, the subjects in (8)
tackle the task systematically by breaking it into two parts; they begin by lo-
cating the correct picture of the rocking chair and dealing with the meaning
of the unknown lexical item in the process. They then work out where this
piece of furniture is to be placed in the matrix picture of the apartment. S1
scaffolds the task for S2 by encouraging her to relate the definitional informa-
tion he supplies to the picture and he then checks whether she has success-
fully accomplished this (You know now?). This kind of scaffolding does not
occur in (6) or (7). In short, (8) demonstrates how the two learners collabo-
rate effectively to achieve their goal and, in so doing, create ideal conditions
for language acquisition.
What exactly are these conditions? They can best be described with refer-
ence to sociocultural theory, which emphasizes that participants respond to
a task by constructing an activity that matches their own goals (Coughlan &
Duff, 1994). Thus, although the task was intended to be identical in all three
conditions (a requirement of experimental research), the activity it generated
turned out to differ in ways we had not predicted. In the premodified and in-
teractionally modified input conditions, the teacher treated the task as a kind
of test, the goal of which was to measure how well the students could under-
stand the directions and learn the new words.7 Thus, the teacher engaged in
relatively little interactive work to help build what Artigal (1992) calls a shared
indexical territory. Even in the input condition that allowed for interaction, the
resulting discourse was highly constrained and even mechanical. For example
in (7) above, the teacher defined rocking chair using the same item (rock) and,
when asked to clarify, simply repeated the same definition. In fact, the teacher
offered exactly the same definitional information as that provided by the pre-
modified direction. It is not so surprising, then, that the results obtained for
the two input conditions were very similar. In contrast, the students treated
the task as a collaborative problem-solving activity, the goal of which was to
help each other to achieve a successful outcome. They arrived at a joint, ne-
gotiated solution to the task and in so doing ensured that the reference of the
new lexical item was fully understood, thus enabling the less expert learner to
acquire it.
In short, we believe that the results of this study support Lantolf and Pav-
lenkos (1995) contention that learning hinges not so much on richness of
input, but crucially on the choices made by individuals as responsible agents
with dispositions to think and act in certain ways rooted in their discursive
histories (p. 116). The choices made by the learners in the modified output
condition created more opportunities for learning the new words than the
choices made by the teacher in the interactionally modified condition.
CONCLUSION
From a sociocultural perspective on language learning, it is not appropriate to

talk of input and output. One has to consider interaction as a totality, a matrix
in which learning is socially constructed. This perspective enables us to deal
with what we perceived as an inherent problem in the design of our study;
although it was possible to contrive conditions that distinguished premodified
and interactionally modified input, it was much more difficult to contrive con-
ditions that narrowly distinguished modified input and modified output. Modi-
fied output does not occur in a vacuumit occurs as a response to input and
the opportunity to interact. Thus, the modified output condition in our study
did not involve the learners in mere production;8 it also involved provid-
ing them with opportunities to talk in pairs in ways they were familiar and
comfortable with. Thus, we cannot say that our study has demonstrated that
modified output works better than modified input. We can only say that inter-
action that provides opportunities for learners to use and negotiate new vo-
cabulary items in dialogically symmetrical discourse seems to create better
conditions for incidental vocabulary acquisition than interaction in teacher-
controlled exchanges that restrict the kind of intermental activity claimed to
foster learning. We also note, however, that even the teacher-controlled ex-
changes in this study were quite successful in promoting vocabulary learning.
NOTES
1. In later posttests, the advantage for the interactionally modified group over the premodified
group disappeared, possibly because the students in the premodified group subsequently engaged
in intentional vocabulary learning.
2. It should be noted, however, that the group that observed other learners interacting showed
greater developmental gains than the group receiving only scripted input.
3. Other research (e.g., Ehrlich, Avery, & Yorio, 1989) has shown that overelaborated input can
also have a negative effect on comprehension.
4. Given that earlier studies had found that learners receiving unmodified input always perform
worse than learners receiving modified input on measures of comprehension and vocabulary acquisi-
tion, the lack of a control group that received unmodified input cannot be considered a serious
threat to the validity of our study.
5. In observations of the task performance, there were no outward signs of learners trying to
memorize the new words.
6. The learners ability to remember the words over time may have been affected by the fact that
they experienced the modified output condition in Posttest 2. However, some subjects did not have
this experience and nonetheless showed high levels of retention in Posttests 4 and 5.
7. Subsequent discussion with the teacher bore out the fact that she perceived the task as part
of an experiment in which she should behave more like a researcher than a teacher.
8. It would have been possible to design a condition that involved only production of the target
items (e.g., by asking the subjects to simply write directions containing the items). However, such a
treatment would not have tapped modified output, as the subjects could have responded quite me-
chanically by simply substituting items in the same sentence frame (e.g., Find the and put it
in the ). In fact, many of the directions the students wrote prior to interacting in pairs were
of this kind.
REFERENCES
Artigal, J. (1992). Some considerations on why a new language is acquired by being used. Interna-
tional Journal of Applied Linguistics, 2, 221240.
Brown, G. (1995). Speakers, listeners and communication. New York: Cambridge University Press.
Coughlan, P., & Duff, P. (1994). Same task, different activities: Analysis of SLA from an activity theory
perspective. In J. Lantolf & G. Appel (Eds.), Vygotskian approaches to second language research
(pp. 173194). Norwood, NJ: Ablex.
Ehrlich, S., Avery, S., & Yorio, C. (1989). Discourse structure and the negotiation of comprehensible
input. Studies in Second Language Acquisition, 11, 397414.
Ellis, R. (1995). Modified input and the acquisition of word meanings. Applied Linguistics, 16, 409441.
Ellis, R., & Heimbach, R. (1997). Bugs and birds: Childrens acquisition of second language vocabu-
lary through interaction. System, 25, 247259.
Ellis, R., Tanaka, Y., & Yamazaki, A. (1994). Classroom interaction, comprehension and the acquisi-
tion of word meanings. Language Learning, 44, 449491.
Krashen, S. (1985). The Input Hypothesis. London: Longman.
Krashen, S. (1994). The input hypothesis and its rivals. In N. Ellis (Ed.), Implicit and explicit learning
of languages (pp. 4577). New York: Academic Press.
Lantolf, P., & Pavlenko, A. (1995). Sociocultural theory and second language acquisition. Annual Re-
view of Applied Linguistics, 15, 108124.
Long, M. (1981). Input, interaction and second language acquisition. In H. Winitz (Ed.), Annals of the
New York Academy of Sciences: Vol. XX. Native language and foreign language acquisition (pp.
259278). New York: New York Academy of Sciences.
Long, M. (1996). The role of linguistic environment in second language acquisition. In W. C. Ritchie &
T. Bhatia (Eds.), Handbook of language acquisition: Vol. 2. Second language acquisition (pp. 413
468). New York: Academic Press.
Loschky, L. (1994). Comprehensible input and second language acquisition: What is the relationship?
Studies in Second Language Acquisition, 16, 303323.
Mackey, A. (1995). Stepping up the pace: Input, interaction and interlanguage development. Unpublished
doctoral dissertation, University of Sydney, Sydney, Australia.
Nobuyoshi, J., & Ellis, R. (1993). Focused communication tasks. English Language Teaching Journal,
47, 203210.
Pica, T. (1994). Research on negotiation: What does it reveal about second-language learning condi-
tions, processes, and outcomes? Language Learning, 44, 493527.
Pica, T., Holliday, L., Lewis, N., & Morgenthaler, L. (1989). Comprehensible output as an outcome of
linguistic demands on the learner. Studies in Second Language Acquisition, 11, 6390.
Pica, T., Young, R., & Doughty, C. (1987). The impact of interaction on comprehension. TESOL Quar-
terly, 21, 737758.
Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehen-
sible output in its development. In S. Gass & C. Madden (Eds.), Input in second language acquisi-
tion (pp. 235252). Rowley, MA: Newbury House.
Swain, M. (1995). Three functions of output in second language learning. In G. Cook & B. Seidhofer
(Eds.), For H. G. Widdowson: Principles and practice in the study of language (pp. 125144). Ox-
ford: Oxford University Press.

The Roles of Modified Input and Output in The Incidental Acquisition of Word Meanings

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Roles of Modified Input and Output in The Incidental Acquisition of Word Meanings

Uploaded by

Copyright:

Available Formats

SSLA, 21, 285301. Printed in the United States of America.

THE ROLES OF MODIFIED INPUT

This article reports an experimental study of the differential effects of

According to the Input Hypothesis (Krashen, 1985), learners acquire an L2

1999 Cambridge University Press 0272-2631/99 $9.50 285

(1) T: Please put the lampshade on the desk.

(2) S1: Please put the comforter on the bed.

researchers have given less attention to identifying a direct impact for

gard to comprehension and acquisition. The tasks that supplied interaction-

1. Comprehension: This was determined by inspecting the subjects completed ma-

Table 1. Descriptive statistics for the

Premodified Interactionally Modified

Table 2. Scheffe test of differences in comprehension scores

Premodified Interactionally Modified

Premodified input 0.35 1.35**

Vocabulary Acquisition (Recognition)

Table 3. Descriptive statistics for the three vocabulary

Premodified input Interactionally Modified output

Table 4. Scheffe test of differences in vocabulary recognition

Premodified Interactionally Modified

Premodified input 0.83 2.02**

Table 5. Scheffe test of differences in vocabulary recognition

Premodified Interactionally Modified

Premodified input 1.23* 2.67**

ANOVA showed these differences to be statistically significant, F = 7.63, p <

Table 6. Scheffe test of differences in vocabulary recognition

Premodified Interactionally Modified

Premodified input 0.59 1.90**

Table 7. Descriptive statistics for the two vocabulary

Premodified Interactionally Modified

Measure PT2 PT4 PT2 PT4 PT2 PT4

Vocabulary Acquisition (Production)

Table 8. Scheffe test of differences in vocabulary recognition

Premodified Interactionally Modified

Premodified input 0.35 1.35**

Table 9. Scheffe test of differences in vocabulary recognition

Premodified Interactionally Modified

Premodified input 0.22 1.35**

From a sociocultural perspective on language learning, it is not appropriate to

You might also like